Shin Watanabe

• Watanabe, Shin; Horn, Geir; Tørresen, Jim & Ellefsen, Kai Olav (2025). Integrating Bilevel Planning and Offline Skill Learning for Enhancing Mobile Manipulation, 2025 IEEE 21st International Conference on Automation Science and Engineering (CASE). Institute of Electrical and Electronics Engineers (IEEE). ISSN 9798331522476. p. 2275–2280. doi: 10.1109/CASE58245.2025.11163881. Full text in Research Archive Show summary

  Solving complex robotic mobile manipulation tasks requires both planning the sequence of skills to execute and learning how to robustly execute each skill. Planning-based approaches such as task and motion planning (TAMP) can help train skills more efficiently through demonstrations, while learning-based approaches such as reinforcement learning (RL) can help plan tasks more quickly through heuristics. This paper presents a novel approach to generalizing mobile manipulation tasks by synergistically combining sampling-based TAMP and value-based RL. The TAMP solver first generates suboptimal demonstration trajectories of a particular skill, from which an offline RL algorithm distills a robust policy and a value function, the latter serving as a skill feasibility classifier. The policy and the classifier are both fed back into the TAMP workflow not only to improve the skill success rate but also to speed up the planner by sampling robot configurations for which the skill is likely to succeed. We evaluate the approach on a simulated block-pushing domain. Re-purposing a byproduct of an offline skill-learning process leads to an integrated planning and learning system that exploits the awareness of its own skill competence.

• Maeda, Ryuichi; Baselizadeh, Adel; Watanabe, Shin; Kurazume, Ryo & Tørresen, Jim (2025). Adaptive Tidying Robots: Learning from Interaction and Observation. In Asfour, Tamim; Ramírez-Amaro, Karinne; Kim, Joohyung & Cheng, Gordon (Ed.), 2025 IEEE/SICE International Symposium on System Integration (SII). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331531614. p. 185–192. doi: 10.1109/sii59315.2025.10871003. Full text in Research Archive
• Watanabe, Shin; Horn, Geir; Tørresen, Jim & Ellefsen, Kai Olav (2024). Improving Robot Skills by Integrating Task and Motion Planning with Learning from Demonstration. In Capi, Genci (Eds.), 2024 6th International Conference on Control and Robotics. IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331518158. doi: 10.1109/iccr64365.2024.10927043. Full text in Research Archive

View all works in NVA

• Watanabe, Shin; Horn, Geir; Tørresen, Jim & Ellefsen, Kai Olav (2025). 2025 IEEE 21st International Conference on Automation Science and Engineering (CASE). Institute of Electrical and Electronics Engineers (IEEE). ISBN 9798331522476. doi: https:/ieeexplore.ieee.org/xpl/conhome/11163731/proceeding. Full text in Research Archive

View all works in NVA

• Tørresen, Jim; Prestes, Edson; Caleb-Solly, Praminda; Weng, Yueh Hsuan & Watanabe, Shin (2025). CASE 2025 Workshop on Ethical Considerations in Robotics and Automation (WECRA). doi: https:/sites.google.com/view/robot-ethics-case2025-workshop. Full text in Research Archive
• Søyseth, Vegard; Bergflødt, Adrian Anderson; Glette, Kyrre; Watanabe, Shin & Otterdijk, Marieke van (2024). Utstilling av roboter på SENT @ Teknisk museum. Full text in Research Archive
• Karlsen, Einar; Tørresen, Jim; Ellefsen, Kai Olav; Badescu, Claudia Andreea; Otterdijk, Marieke van & Watanabe, Shin [Show all 9 contributors for this article] (2023). Møte den nye helsearbeideren. [Journal]. Elektronikk. Full text in Research Archive
• Bjerkan, Mons Eirik; Watanabe, Shin & Tørresen, Jim (2025). Offline Reinforcement Learning through MPC Bootstrapping. Universitetet i Oslo. Full text in Research Archive Show summary

  Efficient, safe, and robust learning for robotic tasks in the real world is more relevant than ever, with new robots or software improvements emerging rapidly. Accordingly, the popularity of \gls{rl}, and offline \gls{rl} has also massively increased over the last years. There are, however, issues training robots for real-life tasks using \gls{rl} with concerns of safety and the amount of data needed. Offline reinforcement learning is a paradigm with the promise of alleviating some of these challenges. Control algorithms such as \gls{mpc} also aim to solve these complex issues. This approach of using non-learning-based control algorithms has long been the most dominant paradigm within control of complex systems. And has only become more relevant with the emergence of real-time optimization, such as real-time \gls{mpc}. To leverage the best of both worlds, this paper proposes integrating offline \gls{rl} with \gls{mpc} by training an offline \gls{rl} policy solely on pre-collected \gls{mpc} trajectories, thereby effectively bootstrapping the \gls{rl} policy training process. The algorithm aims to avoid excessive exploration for faster convergence towards stronger solutions by leveraging the already provided solutions from the \gls{mpc}, while exploiting the characteristics of \gls{rl} to make up for any weaknesses in the \gls{mpc} solutions to make a robust, safe, and data-efficient solution for task-specific robot training. The proposed method is evaluated in a simulated environment and tested on a quadruped robot navigating through a hilly terrain. Testing shows that the proposed method does not measure up to the performance of pure \gls{mpc}. However, the results indicate that the proposed method can be applied to learn how to solve simpler control tasks.

View all works in NVA